AI agent access control
AI agent access control decides which systems an AI agent may reach and which actions it may take there. It is strongest when something other than the agent checks every call.
In a run
Each call is checked against the agent’s access policy and its credential for the job. A call inside both goes through. A call outside either is refused, and the record holds the refusal and what was outside permission.
Our word for it
Can do only what it was allowed to do.
Two questions on every call
Is this action inside the rules your team approved for this agent? Is it inside what this job was given? Ares answers both before the call reaches the tool or the next agent. The agent does not answer for itself.
An example from the sandbox
In the supplier follow-up run, the purchasing agent may read open orders and email the supplier on file. A hidden line in a supplier attachment asks for a drawing to go to a new address. That send is refused, outside permission. The rest of the job goes on: the delivery date change waits for the buyer, and the expected invoice date is updated.
What a refusal looks like
A refusal is not an error message. It is a line on the record with the agent, the job, the action it asked for and why that action was outside permission. Your reviewers read it the same way they read an approval.
Where it stops
- Permissions are as good as the rules you approve. Ares does not decide what an agent should be allowed to do.
- Access control limits what an agent may call. It does not control what runs inside the tool it calls.
Related terms
- Access policyAn access policy is the set of rules for who and what each AI agent may reach, and what it may do there.
- Least privilegeLeast privilege means an account holds only the access it needs for its task, and nothing more.
Tell us about one workflow.
We will say whether an agent should do it, and what the first step costs.
A senior member of the team replies within two business days.